Spectral binaries of brown dwarfs

Screen shot 2013-08-22 at 2.02.36 AMHow do brown dwarfs form? Some theories point to a star-like birth, accreting material from a molecular cloud, while some others point to a planet-like formation from a pre-stellar disk. Either way, the essential mechanisms for brown dwarf formation remain under debate by theorists. Given the astronomical timelines of star formation (1-10 million years), we cannot witness the formation process in action, but we can study its consequences on the statistical properties of the systems created.

Very Low Mass Multiples

While brown dwarfs are mostly found in isolation (~75% of systems), binary and higher order systems provide a unique opportunity to constrain evolutionary models and understand the physical properties of the low mass population as a whole. Observed distributions of the separation, mass ratio and eccentricity of very low mass (VLM) binaries are the standards against which theoretical predictions are tested. If we organize all the VLM binaries by the separation between their components, we see that there is a big drop off at separations smaller than 4 AU (Figure 1). Is this real? 

Projected separation distribution of 122 confirmed VLM binary systems. Systems resolved by direct imaging are shown in green, all other detection methods are shown in blue. Spectral binaries are shown in red.

Projected separation distribution of 122 confirmed VLM binary systems. Systems resolved by direct imaging are shown in green, all other detection methods are shown in blue. Spectral binaries are shown in red.

The most fruitful method to uncover VLM binary systems has been direct imaging with high resolution telescopes (~70% of systems). This method’s inherent bias is its limit in angular resolution. The best angular resolution at using adaptive optics at Keck is ~0.1 arcseconds (1/36,000 of a degree), which for field brown dwarfs at typical distances of 20-40 pc resolves separations of 2-4 AU. This limit coincides with the observed peak in the separation distribution above. The question is then: are binaries at small separations  (<1 AU) rare or simply being missed?

Spectral Binaries

An alternative method of identifying binary systems we have developed is to look for peculiarities in unresolved spectra. These systems, called spectral binaries, have spectra exhibiting absorption features that are inconsistent with any single spectral type, but can be reproduced as the combination of two dissimilar spectra. Specifically, binary systems with late-M/early-L primaries and T dwarf secondaries stand out due to presence of a methane absorption feature at 1.63μm, which is not expected in the hot atmospheres of late-M and L dwarfs (2500-1500K).

Figure 2 shows an example of a spectral binary, 2MASS J1341-3052.  The best matching spectrum in the SpeX Prism Library doesn’t quite fit, showing discrepancies (the “dip”) near 1.63 µm where methane absorption occurs. However, when we make a binary template by adding together properly-scaled L dwarf and T dwarf spectra, we see that this blended-light spectrum is an excellent match. 2MASS J1341-3052 is a strong candidate to be a tight, unresolved binary.

Best fit binary template for the brown dwarf 2MASS 1341-3052. The black line is the source’s spectrum. The red and blue lines are the primary and secondary spectra which are scaled and combined to produce the green line. The gray line is the noise from the candidate spectrum. The inset zooms in around the 1.63um methane absorption feature.

Best fit binary template for the brown dwarf 2MASS 1341-3052. The black line is the source’s spectrum. The red and blue lines are the primary and secondary spectra which are scaled and combined to produce the green line. The gray line is the noise from the candidate spectrum. The inset zooms in around the 1.63um methane absorption feature.

Discovering New Spectral Binaries

In order to look for more such systems, we developed a set of spectral indices that are tuned to the specific peculiarities spectral binaries show (spectral indices are flux ratios that emphasize the contrast between two wavelength ranges). We calculated these indices for ~800 spectra in the SpeX Prism Library, including four independently-confirmed binary systems. These spectral indices are compared against each other, as illustrated in Figure 3. Outliers the are near the known binaries are our candidates.  We then did the same binary template fitting analysis described above to validate these candidates. As a result, we have found fourteen candidate binaries which are now being followed up with high resolution imaging, high resolution spectroscopy and astrometry.

Comparison of two spectral indices showing the binary benchmarks as red stars, spectral binary candidates as green triangles and blue contaminants as blue circles. The region delimited by the benchmarks encloses sources with high probability of being binaries.

Comparison of two spectral indices showing the binary benchmarks as red stars, spectral binary candidates as green triangles and blue contaminants as blue circles. The region delimited by the benchmarks encloses sources with high probability of being binaries.

Contamination

The additional flux in the J band coming from the T dwarf secondary induces a slightly blue J-K color in unresolved binary spectra. As a consequence, objects with even bluer colors, such as subdwarfs and unusually blue L dwarfs, are the main contaminants in our method. The peculiarities in their spectra are due to various levels of low metallicity, thin cloud coverage, or large-grain clouds in their atmospheres, rather than binarity. The rarity and scarcity of these objects in the SpeX Prism Library allowed for better binary fits as compared to single fits, but since the binary fits themselves were poor, these candidates could be easily discarded. 

Are Tight Binaries Common?

To address our earlier question, we compared the small number of spectral binaries with confirmed separations to the separations of the larger population of VLM binaries (see Figure 1). A Kolmogorov-Smirnov test, designed to gauge the level of correlation between two distributions, yielded a low probability (25%) that the samples come from the same distribution. In other words, it appears that there may be more tighter binaries than surmised from direct imaging programs.  However, this is based on small numbers of confirmed spectral binaries; many more will need to be characterized before a significant difference can be confirmed or ruled out.

Future work includes constraining the sample to be volume-limited (up to 25 pc), understanding the selection biases from our technique, and ultimately compiling new separation, mass ratio and eccentricity distributions which include very tightly-separated (<1 AU) binaries.

This research has been published in the Astrophysical Journal, 2014, Vol. 794, pp. 143.

Speak Your Mind

This site uses Akismet to reduce spam. Learn how your comment data is processed.